POISE method infers parent-of-origin effects in standard GWAS data without family pedigrees

Parent-of-origin effects (POEs) — where the phenotypic consequence of an allele depends on whether it was inherited maternally or paternally — are relevant to growth, metabolism, and neurodevelopment, and are a known feature of imprinted genomic regions. Conventional methods for detecting POEs require family-based data in which parental origin of transmitted alleles can be established directly, making large-scale population screens logistically difficult and expensive.

A preprint posted to bioRxiv on 10 June 2026 presents POISE (Spectral Inference of Parent-of-Origin Effects in Unlabeled Genomic Data), a method that infers POEs from standard genome-wide association study (GWAS) datasets in which parental labels are absent. The approach is based on community detection from machine learning, using spectral decomposition to identify statistical signatures consistent with parent-of-origin dependence without requiring inheritance information.

The authors argue that by making POE analysis tractable in the large biobank cohorts already available — where family structure is not routinely recorded — POISE could substantially expand the catalogue of imprinted or parentally biased loci in complex traits. The method has potential applications in studies of imprinting disorders and developmental conditions with known parent-of-origin aetiology. The preprint has not yet been peer-reviewed.